You will fire your analyst in the next 90 days.
You will fire your analyst in the next 90 days.
Not because you have to. Because you'll realize you never needed one.
There are 3 types of founders right now.
And the gap between them is about to become the difference between who scales and who quietly stalls out.
- The Dashboard Hoarders
The ones with 14 tabs open — GA4, Mixpanel, Hotjar, a Notion doc someone built once in 2024 and nobody has touched since.
They check "analytics" every morning like a religious ritual. Coffee first, dashboard second. They screenshot charts for investor updates because it makes the deck look serious. They can tell you their bounce rate to two decimal places, their session duration by device, their funnel drop-off by cohort.
Ask them what to actually change this week, and they go quiet. Then they say "let me dig into the data more." They've been "digging into the data more" for eighteen months.
- The Gut-Feel Operators
The ones who rejected all of it. "Data doesn't matter, just ship and vibe-check it."
They move fast, and I respect that more than the hoarders do. But they're right about the disease and wrong about the cure. Dashboards being theater doesn't mean the alternative is guessing. It means the tooling was never built for a human to actually use.
- The Interpreters
The ones who noticed something the other two never did: the problem was never a lack of data.
It was a lack of translation.
Nobody is short on numbers in 2026. Everybody is short on someone — or something — to tell them what the numbers mean and what to do about it before Friday.
I've shipped a dashboard nobody opened twice. Including me.
Beautiful thing. 40 charts, color-coded, exportable, the kind of screen that looks incredible in a product demo. Bounce rate, session duration, funnel drop-off by stage, cohort retention curves. I was proud of it for about a week.
Then I watched the analytics for my own analytics tool, and the number that mattered was this: average time on the dashboard page, 40 seconds. People opened it, scanned it, and left — because every single chart required them to already know what they were looking for before it told them anything useful.
That's the trap almost every analytics tool in existence sets, quietly, on purpose or not. It hands you the raw feed and calls that "insight." It gives you a library and calls it an answer.
Bookmark this one. It's long, and it will save you the better part of a year I spent building the wrong thing before I understood why it was wrong.
I. Why More Data Made Us Dumber, Not Smarter
Every analytics tool since the early 2000s optimized for the same thing: collect more.
More events, more custom properties, more segments, more charts, more "insights" buried inside reports nobody opens past page one. The pitch was always some version of "now you'll finally understand your users."
Nobody stopped to ask the actual question underneath all of it: does a solo founder, checking a dashboard at 11pm between customer emails, have the training to read a multi-variate funnel chart the way a trained data analyst does?
No. And they were never supposed to need to. That was the entire unspoken promise of hiring a data team in the first place — someone else does the reading so you can do the deciding.
Most founders can't afford that person. So they got handed the analyst's raw materials instead of the analyst's answer, and were told that was progress.
"200 reports. 40 charts. Zero answers." That line isn't a knock on any single tool. It's a description of the entire category, and most people building in it know it's true and ship the 41st chart anyway, because charts are what gets funded.
Jeff Bezos put it simply, years before any of this existed: "If you double the number of experiments you do per year, you're going to double your inventiveness." He was talking about action — running the experiment, learning, moving. Every dashboard on the market today optimizes for the wrong half of that sentence. It maximizes what you can look at. It does nothing to maximize what you actually do.
II. The Interpreter Layer
Here's the model I now run everything through, and the one thing I'd want you to walk away with if you read nothing else in this letter.
There are two layers sitting between "a user does something on your product" and "you make a decision because of it."
Layer 1 is Collection. Every tool on the market does this well now. Pageviews, clicks, sessions, custom events, revenue attribution — this layer is a commodity. Genuinely, none of it is technically hard anymore. Any competent engineer can wire up tracking in an afternoon.
Layer 2 is Interpretation. Almost nobody does this, and it's the only layer that was ever actually valuable. This is the layer that takes 40 charts and compresses them into one sentence: "your error rate spiked 12% after Tuesday's deploy, and it's costing you roughly 3% of signups." Not a chart. Not a dashboard. A sentence you can act on before lunch.
The Dashboard Hoarders live entirely in Layer 1 and mistake it for analysis. They are, functionally, doing unpaid data entry with extra steps and better fonts.
The founders who win the next five years will stop touching Layer 1 directly, the same way nobody hand-codes assembly anymore. AI sits at that layer permanently, instead of you. You live in Layer 2, full time, and that's the only place a founder's judgment was ever actually needed.
If you've ever opened an analytics tool, stared at a chart for ninety seconds, felt a vague sense of "I should probably do something about this," and then closed the tab without doing anything at all — you're stuck in Layer 1. That's not a discipline problem, and it's not a you problem, even though every one of these tools quietly trains you to blame yourself for it. It's a tooling problem, and it has been for twenty years.
III. Which One Are You, Actually
Quick gut check, because I know some of you are reading this thinking it doesn't apply to you.
If you have more than three analytics tabs open right now, and you cannot tell me your one number that matters without opening any of them — you're a Hoarder, whatever you tell yourself.
If your last product decision was made off a feeling and you haven't checked a single number to confirm or kill that feeling within a week of shipping — you're a Gut-Feel Operator, and you're leaving obvious wins on the table purely out of allergy to the word "metrics."
If you've ever built or paid for a dashboard and privately admitted to yourself that you don't actually open it anymore — congratulations, you already know everything in section I. You just haven't fixed it yet.
IV. The One Number Audit
Here's the actual tool. Not advice — a ten-minute exercise. Do it today, not "this week."
Open whatever analytics you currently use and answer these, in order, honestly:
— Without looking at anything, write down the one number that would genuinely worry you if it dropped 20% overnight. Not a vanity metric. The real one. If you can't name it inside five seconds, you don't have a metric, you have a museum you occasionally visit.
— Now open your dashboard and count how many clicks it takes to actually find that number. If it's more than one, your tool wasn't built for you. It was built to look comprehensive in a sales demo, which is a different design goal entirely, and you've been paying for someone else's demo.
— Now ask yourself the question that actually matters: what would you do differently this week if that number moved 15% in either direction? If your honest answer is "I'd probably look at some more charts to figure out why," you don't have an analytics problem. You have an interpretation problem, and no new tool fixes that until the question comes before the chart, not after it.
— Last one, and it's the one people skip: who currently reads this number besides you? If the answer is nobody, you've built a private ritual, not a decision system. Fix that before you fix anything else on this list.
Most founders fail on step one. Not because they're bad operators — because every tool they've ever touched trained them to admire data instead of interrogate it. That's a training problem built into the software, and it's not your fault, but it is now your job to unlearn it.
I got tired of failing my own audit, which is most of why I ended up building an AI layer that just hands me the sentence instead of the chart in the first place. Small, free-to-start, cookieless thing — happy to share what Layer 2 looks like in practice if anyone wants to see it, no pressure either way.
V. The Shift Nobody's Naming Yet
The scribe became the editor. The hand-weaver became the machine operator. The bookkeeper became the person who reads the P&L instead of the person who builds it line by line. Every single time a mechanical skill gets automated, the human role doesn't disappear — it moves up one layer, and it always has.
Analytics is having its scribe-to-editor moment right now, in real time, and almost nobody in the space is naming it that directly, because half the industry still makes its money selling you the scribe's tools.
You don't need to read charts anymore. That skill is being automated the same way manual bookkeeping was. What you need instead is the judgment to recognize a good answer when your AI hands you one, and the speed to act on it before a competitor running the same playbook gets there first.
The founders who win won't be the ones with the most dashboards open, or the fanciest funnel visualizations in their pitch deck.
They'll be the ones with none open at all — because their data finally learned how to talk, and they finally stopped treating a museum like it was a nervous system.